A voiceprint is a biometric feature of a human being that can be used to uniquely determine a person's identity. Voiceprint recognition can be based on anthrophonic and behavioral features of a particular human speaker. Security provided by voiceprint authentication can be comparable to other biometric authentication technologies, such as fingerprint or iris authentication. The hardware for performing voiceprint recognition includes a microphone, which is relatively inexpensive. Most voiceprint recognition methods are unrelated to language, dialect, or accent of a speaker, and there is normally no privacy issues in using voiceprint recognition. Voice signals are also convenient for transmission and reception, which can be suitable for applications based on the Internet or other telecommunication networks.
Voiceprint recognition can include two stages. The first stage is a registration stage, where a user records one or more voice segments to a voice recognition system. The system can construct a voiceprint recognition model for the user by using voiceprint features extracted from the recorded voice segments. The second stage is an application stage, where the user reads a predetermined text provided by the voice recognition system. The system automatically records the voice of the user, extracts voiceprint features, conducts a matching operation between the extracted voiceprint features and the constructed voiceprint recognition model, and determines whether the voiceprint features match the voiceprint recognition model for user identity authentication.
Voiceprint recognition can include text-dependent recognition and text-independent recognition. Text-dependent voiceprint recognition requires a user to read a piece of predetermined text for voiceprint registration. Voiceprint recognition modeling can be more easily performed since all users using text-dependent recognition read the same standard text. In contrast, text-independent recognition does not require a user to read a piece of specific text, which makes the voiceprint recognition modeling more difficult since no standard text can be used as a reference to register the user's voiceprint.